Computational analysis of kinase inhibitor selectivity using structural knowledge.

Area of Science:

• Biochemistry and Medicinal Chemistry
• Computational Biology and Cheminformatics

Background:

• Kinases are crucial in disease signaling pathways, making kinase inhibitor selectivity a key concern for drug design and safety.
• Existing methods for assessing kinase selectivity are often costly, limited, or computationally intensive, hindering large-scale profiling.

Purpose of the Study:

• To introduce the KinomeFEATURE database for efficient kinase binding site similarity searches.
• To develop a computational approach for accurate, large-scale kinase inhibitor selectivity profiling.

Main Methods:

• Utilized physiochemical descriptors to characterize protein microenvironments for kinase binding site similarity searches.
• Developed and applied the KinomeFEATURE database for kinome-wide selectivity profiling of kinase inhibitors and drug candidates.

Main Results:

• Achieved >90% accuracy in initial selectivity predictions for known kinase inhibitors.
• Identified 55 promiscuous kinases and unexpected cross-activities, such as between PKR and FGFR2.
• Predicted novel and validated off-targets for kinase drug candidates, elucidating mechanisms of cross-activity.

Conclusions:

• The KinomeFEATURE database offers a powerful tool for large-scale kinase inhibitor selectivity profiling.
• This approach can significantly contribute to advancing kinase drug development and improving clinical safety assessments.